Accuracy

Unnecessary insults directed at people who've at the very least displayed a willingness to help FTW!

[ QUOTE ]
Over the 4 day weekend, I left my computer running with Eviscerate on Autoattack. (Both my target and my character were set to not take damage so there was no worry about killing each other.)

I attacked a total of 38917 times.
I hit 29275 times.
I missed 9642 times.

That is a 75.22% overall to hit chance against an even level critter with no defense powers.

There is no "Accuracy Wrap Around" -- otherwise Blasters with Targetting Drone + Aim + Build Up would never, ever hit, which is not the case.

[/ QUOTE ]

Castle:

I have never doubted the total over a long, long term like this.

But what about RUNS? That is... if you had 29,000 hits in a row followed by 9,000 misses in a row, over your total, you still have 75% accuracy but clearly something would be wrong.

People have noticed things like, "When I miss, I always miss TWICE in a row..." or "When I get hit, I always get hit THREE times in a row, never once, twice, or four times" and this sort of thing.

I think there is a bug that is making RUNS happen, but not affecting the overall percentages (probably because runs of hits and misses are appropriately likely, so a long to-hit run may be likely 75% more often than a long miss run). What players notice is not long-term averages but unusual runs. For example tonight, I missed 6 times in a row with my tanker, who has an SO in every attack, invincibility, AND was using build-up. She should be at least 90% or so to hit in such cases (vs. yellow minions) and so 6 misses in a row seems very odd. When this was followed by 8 out of 11 attacks ALSO missing (each one at least 2 in a row) it starts to look suspicious. Admittedly this is a very tiny sample size, but is there any chance you guys could log not just the NUMBER of hits and misses but also the sequence? It would be useful to have this and test the sequence, to see if the length and frequency of runs is greater than one would expect due to random chance alone.

I would be happy to do the appropriate inferential statistics for you if you can provide me with the sequence. I just don't have the means to track a long enough sequence to obtain the data myself.

F

I am a lvl 20 SS/EA Brute. Although I do not have any fancy numbers to back up my claims(surely I'll be disregarded), my accuracy is pathetic. I have 2 DO acc enhancements in Jab, Punch and Knockout Blow and 3 DO acc enhancements in Haymaker. I miss enough to make the experience frustrating at the very least and downright unenjoyable on other occasions. I realize SO's will help, but my influence and current level doesnt allow for that yet. THIS NEEDS TO BE LOOKED AT!! ITS JUST NOT FUN! Thx for listening.

i acctually noticed this too. The runs seem to come in at weird intervals too(some very unopertune but hay). I missed AS with build up, then soaring dragon, then Gamblers cut, and one more with brawl then i hit for 9 in a row...its just a little odd some times.

[ QUOTE ]
Unnecessary insults directed at people who've at the very least displayed a willingness to help FTW!

[/ QUOTE ]

Insults can be quite necessary, Obi-Not the one...

Double posts just ruin it, don't they.

[ QUOTE ]
Over the 4 day weekend, I left my computer running with Eviscerate on Autoattack. (Both my target and my character were set to not take damage so there was no worry about killing each other.)

I attacked a total of 38917 times.
I hit 29275 times.
I missed 9642 times.

That is a 75.22% overall to hit chance against an even level critter with no defense powers.

There is no "Accuracy Wrap Around" -- otherwise Blasters with Targetting Drone + Aim + Build Up would never, ever hit, which is not the case.

[/ QUOTE ]
How many streaks of missing were there, say missing up to and over 5 times in a row?

Continuing with my assumption that it's not Accuracy that is bugged, but Defense, I've been saving my HeroStats files after fighting each enemy type.

I am a level 48 Dark/Dark Scrapper. I have two level 50 Accuracy SO's in Air Superiority, Smite, and Midnight Grasp (HeroStats doesn't track Shadow Maul due to the obscure 'hit' messages.). This should give me a base of 130% Accuracy according to CoHPlanner.

Fighting Werewolves and Hydra at +2 levels (all minions, these are Portal missions), my Accuracy for all attacks was 95% +/- a small margin of error.

Fighting Carnie minions at +2 levels (mostly, very few +3's tossed in) my Accuracy for all attacks was 80% +/- a small error.

Let's say that I lose 10% Accuracy per level difference, I should be at 110% when fighting +2 minions. Do Carnies have a 30% Defense? Is this working as intended?

I have another session where I fought nothing but +2 Devouring Earth Boulders. I was level 46 with 2 level 45 SO's in the attacks (120% Accuracy). In that case I had 65% with Air Superiority (all Smashing), 56% with Midnight Grasp (all Negative Energy), and 71% with Smite (half Smashing, half Negative). I don't know what's up with that.

This is where the developers have an edge. They know what Defense Carnies, Boulders, and Werewolves have. They know what the Accuracy penalty is for higher level enemies. All we can do is report numbers from the LIVE server.

What do you guys think?

I recently found my Quality Control Statistics textbook from a class I taught once, and I'm thinking I can try applying the techniques it contains to see if the hit/miss is appropriately random.

But I don't have HeroStats. Any recommendations as to how to get appropriate input for this process?

It'd be nice if I could get the game to log the calculated expected percentage to hit for each attack. I wonder if that's in the logs Cryptic keeps? it'd be meaningless to most players but they could put in an option for something like "verbose logging" hmm...

[ QUOTE ]
I recently found my Quality Control Statistics textbook from a class I taught once, and I'm thinking I can try applying the techniques it contains to see if the hit/miss is appropriately random.

But I don't have HeroStats. Any recommendations as to how to get appropriate input for this process?

[/ QUOTE ]

HeroStats is nice and free, if it works for you.

As I said, it's my observation that my own personal hit/miss issue is completely dependent upon the enemy that I am fighting. More appropriately, "Sometimes I just miss constantly."

[ QUOTE ]
It'd be nice if I could get the game to log the calculated expected percentage to hit for each attack. I wonder if that's in the logs Cryptic keeps? it'd be meaningless to most players but they could put in an option for something like "verbose logging" hmm...

[/ QUOTE ]

I'm sure that Cryptic could get that data if they chose to. HeroStats cannot since the chat messages it uses ("Midnight Grasp missed!") don't tell you enough about the target. I guess it could tell you how many times you hit a certain mob type, but it would be up to you to calculate the misses and what level they were against.

It's interesting, to say the least.

[ QUOTE ]
Let's say that I lose 10% Accuracy per level difference, I should be at 110% when fighting +2 minions. Do Carnies have a 30% Defense? Is this working as intended?

[/ QUOTE ]
They might - as a Dark scrapper, you use Negative Energy on all your attacks. If Carnies have defense to negative energy, then that would explain what you're seeing. I have no idea if they really are supposed to have such a defense or not - just saying it would account for the data if they did.

[ QUOTE ]
(Both my target and my character were set to not take damage so there was no worry about killing each other.)


[/ QUOTE ]

*sends _Castle_ a nice bottle of Makers Mark to 'accidently' set my characters to this setting for...(ahem) testing purposes.*

[ QUOTE ]
Let's say that I lose 10% Accuracy per level difference, I should be at 110% when fighting +2 minions. Do Carnies have a 30% Defense? Is this working as intended?

[/ QUOTE ]

Doesn't level difference affect base To Hit which is actually figured in before accuracy enhancements?

[ QUOTE ]


There is no "Accuracy Wrap Around" -- otherwise Blasters with Targetting Drone + Aim + Build Up would never, ever hit, which is not the case.

[/ QUOTE ]

I'd like to point out, what's referred to a "wrap around" doesn't drop a number to zero. It hits some cap, then starts over again, resulting in an arbitrary number. I'm not saying the game does this, just that a wrap around won't consistently floor your hit chance. When I had target drone, at times it did make me suspicious though....



Back when I played WoW, they had perenial bugs wherein debuffs wouldn't always unapply themselves when done. People used this to debuff their stats lower than zero, wrapping it around to extremely high levels. (as I understand it the variable was unsigned, so -1 ended up being the maximum value)

I'm well aware of how wrap-around works. There were several arcade games back in the '80s I had problems with this in as far as score went.

As for 'Streaks' any true random number generator exhibits streaks. A good write up with some helpful visual info can be found at http://bcs.whfreeman.com/ips4e/suppl...cel/04_IPS.pdf

I have to ask. Was this done on a test server or a live one?

Why do I get streaks of 10 misses against blue cons? I should not have streaks like that more than once a year, but I get them 3 and 4 times a day. I feel that the issue is with the Streakbreaker code as well, not necessarily with accuracy overall.

Ten misses against blue cons? Are you sure? Never, ever, ever seen anything approaching that. I know that three misses in a row feels like ten against whites or lower, but come on...seriously?

[ QUOTE ]
I'm well aware of how wrap-around works. There were several arcade games back in the '80s I had problems with this in as far as score went.

As for 'Streaks' any true random number generator exhibits streaks. A good write up with some helpful visual info can be found at http://bcs.whfreeman.com/ips4e/suppl...cel/04_IPS.pdf

[/ QUOTE ]

Which is why some random song players take steps to make themselves actually less random. People don't like long streaks in their random numbers. It's not fun.

[ QUOTE ]
I have to ask. Was this done on a test server or a live one?

[/ QUOTE ]

From reading what he did as a test, leaving a guy on auto attack, attacking something, for 4 DAYS, without killing it... you can't DO that on live, so it HAD to be an internal test server. Am i right? You can't do crap like that on a live, or even OUR test server.

I don't think the problem some of us are noticing has anything to do with what happens when an attack actually works. I think accuracy from a static I stand here and hit you works fine.

The problem I'm noticing is attacks don't always register. Siren's Call and Howl is the effect I'm really noticing. About 1/50 (maybe less, I'd guess once every 3-4 missions) or so the attack animation happens but the attack itself DOES NOT.

I've really started to notice it with Howl and Siren's Call mostly because if the attack doesn't fire (on the server) the sound effect doesn't either. Those sound effects themselves are particularly unique so easy to notice. I'm actually starting to be able to predict mass misses with those attacks and I'm usually running before the "missed" word even apears. I know when I'm going to miss everything.

When the attack doesn't fire every mob it would have hit gets a "missed" above his head (usually after a longer than normal delay... I think... there is a delay for a normal "missed" delay anyway... ie damage numbers float up and one beat later the missed words appear).

So what I see is... move into position, fire Howl etc. then I see the Howl animation but no sound effect. Then I see 3-6 "missed" floating above the enemy heads. The recharge timer has fired locally, so I can't reuse the attack.

I really think there is a disconnect between what we see on our client and what the server is doing. And that some effects in some play styles makes it apear that accuracy is a problem because the server isn't acknowledging attacks under some circumstances.

There probably is no problem with the accuracy code, but that doesn't mean we are hitting as much as we should. I don't think you'd notice the missed sound effect with melee attacks etc. Simply too much going on, so all you'll notice is that you seem to be missing more than you think you should. But when you rely on an effect like Siren's Call you really start to notice that it doesn't always fire.

[ QUOTE ]
[ QUOTE ]
I have to ask. Was this done on a test server or a live one?

[/ QUOTE ]

From reading what he did as a test, leaving a guy on auto attack, attacking something, for 4 DAYS, without killing it... you can't DO that on live, so it HAD to be an internal test server. Am i right? You can't do crap like that on a live, or even OUR test server.

[/ QUOTE ]

Certainly not with the daily maintenance schedule.

[ QUOTE ]
[ QUOTE ]
I have to ask. Was this done on a test server or a live one?

[/ QUOTE ]

From reading what he did as a test, leaving a guy on auto attack, attacking something, for 4 DAYS, without killing it... you can't DO that on live, so it HAD to be an internal test server. Am i right? You can't do crap like that on a live, or even OUR test server.

[/ QUOTE ]

I tend to agree which is why I would also say it proves nothing. Have we not seen how reliable those test servers are already?

[ QUOTE ]
Ok, that's nice...what enhances did you have in your attack? Did you get a 75% with an accuracy SO in it? If so that's not such a happy thing...

[/ QUOTE ]

How about we assume _CASTLE_ knows his job, please?

[ QUOTE ]
Why do I get streaks of 10 misses against blue cons? I should not have streaks like that more than once a year, but I get them 3 and 4 times a day. I feel that the issue is with the Streakbreaker code as well, not necessarily with accuracy overall.

[/ QUOTE ]

Apologies if this has been brought up earlier in this thread, but there's a psychological factor at work here, too: we're much, much more likely to notice streaks of 10 misses than equally unlikely streaks of hits.

Of course, assuming accuracy is 75% (I know it's higher vs. blue minions, but I forget how much and I'm lazy), the probability of 10 misses in a row is 9.5367x10^(-7), which is approximately the same as getting 48 hits in a row. This is so unlikely that you'd have to execute something on the order of one million attacks before you'd expect to see it happen even once. If you're seeing this consistently, then something is seriously broken -- but to such a degree that I'd be inclined to discount your claims as hyperbole without further evidence (demorecords, logs, etc.) if I were Castle or Geko.